pandas groupby 滚动不均匀时间 [英] pandas groupby rolling uneven time

问题描述

我在熊猫滚动时遇到了一些麻烦.这是我的数据集的简化版本:

I am having some trouble with pandas rolling. Here a simplify version of my dataset:

df2 = pd.DataFrame({  
    'A' : pd.Categorical(["test","train","test","train",'train','hello']), 
    'B' : (pd.Timestamp('2013-01-02 00:00:05'),
                   pd.Timestamp('2013-01-02 00:00:10'),
                   pd.Timestamp('2013-01-02 00:00:09'),
                   pd.Timestamp('2013-01-02 00:01:05'),
                   pd.Timestamp('2013-01-02 00:01:25'),
                   pd.Timestamp('2013-01-02 00:02:05')),
         'C' : 1.}).sort_values('A').reset_index(drop=True)
>>> df2
       A                   B    C
0  hello 2013-01-02 00:02:05  1.0
1   test 2013-01-02 00:00:05  1.0
2   test 2013-01-02 00:00:09  1.0
3  train 2013-01-02 00:00:10  1.0
4  train 2013-01-02 00:01:05  1.0
5  train 2013-01-02 00:01:25  1.0

我想要一个 10 秒的滚动窗口，以获得以下输出:

I would like to have a rolling window of 10s, to get the following output:

       A  count
0  hello   1
1   test   2
3  train   1

我尝试使用 groupby 和滚动.

I try the groupby and rolling.

df2.groupby('A').rolling('10s', on='B', closed='right').C.sum() 

我从过去的10 年代"观察中得到了滚动窗口，这不是我想要的:

I get the rolling windows from the past '10s' observation, which is not what i am looking for:

A      B                  
hello  2013-01-02 00:02:05    1.0
test   2013-01-02 00:00:05    1.0
       2013-01-02 00:00:09    2.0
train  2013-01-02 00:00:10    1.0
       2013-01-02 00:01:05    1.0
       2013-01-02 00:01:25    1.0 

我也尝试重新采样，但我无法得到结果.

I also try resampling, but I am not able to get the result.

grouped = df3.set_index('B').groupby('A').resample('S' )['C'].count()
grouped.reset_index().groupby('A').rolling(window=10,on='B' , min_periods=1).sum() 

推荐答案

我认为你必须试试这个:

I think you have to try this:

df2.groupby('A').rolling('11s', on='B').agg({'C': 'sum'}).groupby('A').max()

这篇关于pandas groupby 滚动不均匀时间的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！